Consider a sentence, “Population of India is 1.3 billion”. The sentence is expressing a numerical fact (population = 1.3 biliion) about a country (India). Equivalently, the interesting information contained in the sentence can be expressed as a 3-tuple (India, Population, 1.3 billion). The goal of our project was to come up with a extractors that can take such a sentence and return the corresponding 3-tuple.
We came up with two solutions to the problem: a rule based system called NumberRule, and a graphical model called NumberTron.